1. Field of the Invention
The present invention relates to a technology for generating inspection logic used in a board inspection apparatus, particularly to the technology in which the inspection logic can be generated from a small amount of teaching data.
2. Description of the Related Art
Conventionally, there has been proposed a board inspection apparatus for inspecting surface-mounted quality of a printed board on which many electronic components are mounted, particularly the board inspection apparatus for performing the inspection by using an image in which the board or a part of the printed board is taken. In such pieces of inspection apparatus, it is necessary that inspection logic be previously established to determine whether the surface-mounted quality of the components is acceptable or defective.
Sometimes the following technique is adopted in the inspection apparatus in which the image is utilized. That is, an image in which an acceptable sample is taken and an image in which a defective sample is taken (teaching data) are analyzed to determine a difference between the acceptable image and the defective image, and the resulting inspection logic is used when analyzing an image to determine whether the surface-mounted quality of the corresponding component is acceptable or defective.
In this case, in order to generate the high-reliability inspection logic, it is desirable to previously prepare a sufficient amount of teaching data. However, at the actual production site, it is difficult to prepare a sufficient amount of teaching data before a new component is brought for inspection. Particularly, for a defective product, it is difficult to predict which kind of defect is generated during production, and there are innumerable variations in defects. Therefore, it is very difficult to prepare the sufficient number of samples that cover all the defects.
Accordingly, a sufficient amount of teaching data cannot be prepared for a certain time after the new component is brought for inspection, which results in a problem that inspection logic reliability cannot be improved.
Thus, there remains a need for a method for generating highly reliable inspection logic for a new component, which allows automation of teaching while requiring a reduced amount of teaching even in cases where there is a small number of samples.